Botelho, Lucas2023-09-272023-09-272023-09-272023-09-25http://hdl.handle.net/10012/19968In this research, a monitoring system for thermal processes was developed which measures the most critical process phenomena, such as thermal dynamics (peak temperature, heating rate, and cooling rate) and geometric features, in real-time, which can be used for quality assurance and real-time feedback control. Thermal processes are a subset of manufacturing processes that are characterized by heating materials with a concentrated heat source to alter the properties of the materials or join them. Metal additive manufacturing and arc welding processes are considered thermal processes, where the concentrated energy source may be in the form of a laser, electron beam, electric arc, etc. While thermal processes can be used to create complex components without the limitations of traditional manufacturing, process disturbances may cause deviations from expected results. During thermal processing, geometry and thermal dynamics of the heat affected zone (HAZ) directly influence the quality of the produced products. Therefore, it is critical to have an accurate tool to monitor the geometry and thermal dynamics in real time to better assure the quality of the part. Various sensors are available to measure these properties, though imaging is a common theme among thermal process monitoring. Imaging is an effective technique since it allows for non-contact in-situ measurements. Imaging in different wavelengths can provide different information regarding the HAZ, such as the temperature distribution from infrared (IR) light. While high resolution, and high frame rate geometry measurements from visible light can be monitored directly. Moreover, processing images with machine learning algorithms has also been shown to be capable of predicting porosity and detecting defects in the part being manufactured. Therefore, the monitoring system designed in this research features high dynamic range (HDR) visible light and IR dual camera sensors with a common optical path to monitor the geometry and thermal dynamics, with the potential to implement machine learning to monitor other features in the future. An enclosure was designed to house both sensors with a common optical setup for the sensors to have a similar field of view (FOV). In this work, the IR sensor was used to create a dataset to predict the temperature distribution of the HAZ with the HDR sensor. From the temperature distribution, thermal dynamics such as peak temperature, cooling rate, heating rate, solidification time, and melting time were calculated in real-time to estimate the material properties of the final part. The HDR sensor was also used to predict the geometry of the deposited material (clad). Using the same sensors, the height and width of the deposition are estimated from the captured images in real-time which are used for deposition geometry control. The geometry prediction algorithm evolved during this work with different algorithms and features used in the measurements to improve the robustness and accuracy of geometry measurements. To test the effectiveness of the monitoring system, laser heat treatment (LHT) experiments were conducted to initially validate the thermal dynamics measurements. Thermal dynamics were then further validated during laser directed energy deposition (LDED), which was additionally used to validate the geometry measurements of the clad. Moreover, gas metal arc welding (GMAW) experiments were conducted as well to demonstrate the potential for using this system for different energy sources and materials. The developed dual sensor camera was shown to be capable of capturing images in real-time during thermal processes. Processing the visible-light images allows the geometry of the HAZ to be monitored, while the IR sensor provides its temperature distribution. The system was shown to be robust enough to capture data with multiple materials (stainless steel and nickel-based alloys) and with different energy sources (laser and electric arc). The thermal dynamics measured with this tool have been shown to correlate to the material properties of the produced parts, thus demonstrating the potential to infer the material properties from these measurements. It has also been shown that a cost-effective alternative design using the visible light sensor to predict the temperature distribution with calibrated measurements from a pyrometer may be used for temperature measurements in thermal processes. Therefore, the developed monitoring system is shown to be an effective monitoring and control tool for various thermal processes.enadditive manufacturingdirected energy depositiongas metal arc weldingreal-time monitoringconvolutional neural networkmachine learningDoctoral Thesis